Shop OBEX P1 Docs P2 Docs Learn Events
microSD board for your Propeller for <$15 (with a free Propeller chip)... An idea — Parallax Forums

microSD board for your Propeller for <$15 (with a free Propeller chip)... An idea

Cluso99Cluso99 Posts: 18,069
edited 2012-07-30 00:29 in Propeller 1
I am considering making a microSD add-on board with a difference...
  • ~0.8"x0.7" (~20x17.5mm) pcb smt
  • microSD socket
  • Propeller chip
  • Xtal (non-socketed?) I think 6MHz because they are cheapest
  • 24LC64 (8KB) eeprom with SD boot code
  • 1/2/3/4 wire connection to your propeller board (various protocols)
  • requires 3V3 and GND (i.e. no regulator on the pcb)
Purpose...
To provide fast and buffered connection to the microSD card. The included prop can read ahead and write behind, compare before writing to save wearing out a card if write is not required.

Interface...
1pin for simple and therefore slow interface (where your prop doesn't have pins available)
2pin high speed serial or lower speed I2C
3pin high speed with clocking
4pin - just look like a SPI chip

I am not bringing out other pins from the prop so please don't ask (saves pcb space because the pcb cost becomes an issue).
There is no need for a PropPlug connection as the 24LC64 will be preprogrammed to boot from the microSD.

Assembled cost would be =<US$15 plus unregistered postage.
Any interest or comments???

Comments

  • Heater.Heater. Posts: 21,230
    edited 2012-06-17 21:47
    Not sure I see the point here.
    Most card reading is not going to be any quicker for having read buffering. How often are you reading the same few kilobytes over and over.

    Edit: I suppose sector read ahead would help though.

    Write buffering always worries me. You think you have written something but it is not written yet.
    As for reducing wear, I'm not sure how often one is likely to be writing the same data to the same sectors,would any data comparing help much. Besides SD card do their own wear leveling.

    However such a card may make sense if you move the entire FAT file system into it. That would save all that code space in the "host" prop.
  • Cluso99Cluso99 Posts: 18,069
    edited 2012-06-17 22:15
    Often you are accessing successive sectors. When loading code into the prop, there can be up to 32KB, or 2KB if its for a cog. Now, it mak be posible to bypass hub completely on the main prop, and this could save some time to.

    I agree, moving the FAT would help a lot.

    An OS could identify the objects to be held in hub ready for access.

    Postedit
    Another possibility is accessing the card in the nibble mode. It is much faster!!!
  • TubularTubular Posts: 4,706
    edited 2012-06-17 22:25
    Sounds interesting. If you have at least 3gpio I could use it for logging from TSL3301 line scan sensor, and I do need something small for that.

    There are a couple of similar products out there - eg 4D's uDrive. Not prop based, obviously.

    Not sure if nibble mode would require the full SD license

    Cannot complain about the price - put me down for 2 or 3 if you do go ahead !
  • Cluso99Cluso99 Posts: 18,069
    edited 2012-06-18 00:30
    tubular:
    SD license.. I have read that linux has open source drivers but I have not found them.

    I will probably have a single row of pins so 1x7 or 1x8 less 2 for 3V3 & GND. This will be the interface to the other prop, so there will be 5 or 6 prop pins available in total, but they might not include P30/P31. I am aiming to have the pcb just bigger than the microSD socket + xtal + interface pins. The prop will be on the other side.
  • msrobotsmsrobots Posts: 3,709
    edited 2012-07-07 21:48
    Cluso99,

    do it right!

    Variation A:

    two sd-cards and build a stripe-set. Even and odd Sectors. You can double the speed.

    Variation B

    three sd-cards and use them as RAID.

    Variation C

    go for it! 6 sd-cards having a striped RAID-set!

    still 4 pins left for interface to prop.

    and the connected prop can use it as blockdevice not knowing anything about more then one sd-card.

    I would buy some ..

    Enjoy!

    Mike
  • 4x5n4x5n Posts: 745
    edited 2012-07-07 21:54
    I would buy a few.
  • Heater.Heater. Posts: 21,230
    edited 2012-07-07 22:54
    msrobots,

    That would be RAISD.

    A wonderfully crazy idea.
  • msrobotsmsrobots Posts: 3,709
    edited 2012-07-07 23:55
    yes RAISD not just grown up...

    Enjoy!

    Mike
  • MacTuxLinMacTuxLin Posts: 821
    edited 2012-07-08 00:27
  • pik33pik33 Posts: 2,397
    edited 2012-07-08 08:45
    Good idea. Unbuffered read/write with FAT drivers is slow and latency is unpredicted. I use 8 GB card with Kye's driver and sector read can take 2.5 ms.. .but it can take over 4 ms, when cluster changes, because the driver has to read two sectors: one from fat, one from file. So buffered SD board with nibble mode and fat32 driver with fat buffering and file read ahead will be a good option.

    SPI mode is simply too slow. Playing a wav file is near the border with this 2.5 ms for 512 bytes.
  • KyeKye Posts: 2,200
    edited 2012-07-08 13:56
    FSRW is much faster than my driver. I do not use multi block mode.

    ---

    You have to use SPI mode. The CRC's needed for 4 bit SD mode will choke the prop chip.
  • msrobotsmsrobots Posts: 3,709
    edited 2012-07-08 14:17
    @Cluso99!,

    forgot to answer your first question... YES I would be interested in your one-card-version also. But please try to give access to he other pins. just put pads there. If needed anybody should be able to drill the holes by themself.

    @MacTuxLin,

    nice one. Hm. 4. RAISD with one spare? Or double stripe set! 4 times the speed possible... Any way to get to the other pins?

    Do you sell it assembled? I started soldering after my eyesight is leaving me so even simple things like the parallax IO-board are a challenge to me, yet. But at least that one I finished nicely - way to go for surface mount stuff.

    Enjoy!

    Mike
  • Cluso99Cluso99 Posts: 18,069
    edited 2012-07-08 18:22
    msrobots + anyone interested in solderig smt:
    It's not as difficult as first imagined. It takes special pcb layout (extended pads) and good glasses. I use an attached pair of jewellers eyeglasses which go over my normal glasses. While you have to get the distance accurate to focus, its quite easy. I just use thin solder and a flux pen plus a temperature controlled iron with a fine point. Have some desoldering braid handy, but after a while you rarely require this. I hand solder all the boards I make. The smallest smt parts are the resistor networks (4 resistors) Digikey YC164J series (1206/0603)
  • msrobotsmsrobots Posts: 3,709
    edited 2012-07-30 00:29
    @Cluso99,

    any progress there ? I am still interested in them ...

    btw ... attached FullDuplexSerial_rr004-Autodoc ... with demos of org. FullDuplexSerial and Phils Autodoc comments.

    Enoy!

    Mike
Sign In or Register to comment.